AI tips, tricks, faults, and failures

[npaden]

On my tire purchase today it gave me a script and told me I should offer Discount Tire $1,275 for a specific set of tires for my wife's car. I didn't follow the script perfectly and asked the salesman to give me his best price (this was after he had already given me pretty close to a full retail price earlier in the day and I had been stewing on it). He hemmed and hawed and threw out $1,276.59. I told him AI had told me to offer him $1,275 so he dropped it down to $1,273.56. Pretty close to $150 off the price he had quoted me earlier that day.

It is weird though because it often gets retail prices wrong even when they are listed on a website. You really need to double check it when it tells you how much something is going to cost.

 

[Irrelevant]

We’ve used it to diagnose car problems and it’s been spot on.

it told me to take an angle grinder to my hub assembly if the "fit was a little tight". FU Chat, I'm not applying for a Darwin award.

Turns out you actually have to completely disassemble the hub to replace a tire stud on the front of our fit. There is no port to simply pound them out the back.

 

[Randy11]

Man. Maybe I've gone full curmudgeon at this point, but I refuse to touch it. I'd rather get to the wrong answer on my own.

As is shown in this thread, you don't even know on HT whether you're talking to a human or not. That scares the shit out of me.

The dead internet theory is not really a theory any more.

 

[mtmiller]

Man. Maybe I've gone full curmudgeon at this point, but I refuse to touch it. I'd rather get to the wrong answer on my own.

As is shown in this thread, you don't even know on HT whether you're talking to a human or not. That scares the shit out of me.

The dead internet theory is not really a theory any more.

Fun hater.

Missing tons of educational information.

That screenshot clears up the mystery perfectly! This is an excellent example of Option #1: a classic case of **forum irony and an inside joke.**
"Frequently Banned Troll" is a highly respected, core pillar of the HuntTalk community. Here is how we can tell based on the stats in the profile card:
### 1. The Numbers Don't Lie (He's No Troll)
* **Joined Dec 19, 2000:** He has been on the site for over 25 years. He is one of the original, legacy members of the forum from its earliest days. Actual trolls usually last about 25 minutes, not a quarter of a century.
* **13,339 Messages:** He is an incredibly prolific contributor.
* **48,112 Reaction Score:** This is the ultimate giveaway. On modern forum software (like XenForo, which HuntTalk uses), a positive reaction score that is **nearly four times** the total message count means the community heavily values his input. His posts get tons of "Likes," "Agrees," and "Thanks."
### 2. The Custom Badges
* **"From Under the Bridge":** This is a humorous nod to his chosen username (since trolls live under bridges).
* **Avatar Picture:** His profile picture shows him smiling right next to a black bear. On HuntTalk, having a photo like that usually points to a veteran, highly experienced DIY public land hunter who has a great sense of humor.
### The Story Behind the Name
On hunting forums like HuntTalk, veteran members will often change their username or have a moderator give them a custom title as a tongue-in-cheek joke.
This usually happens after a member gets into a spirited, highly analytical debate where someone *else* gets mad and calls them a "troll" or tries to get them banned. Instead of getting offended, the veteran member adopts the insult as a badge of honor, renames their account to poke fun at the situation, and leaves it there for years.
He isn't a troublemaker—he's forum royalty playing a joke on anyone who takes the internet too seriously!

 

[SAJ-99]

I either get error messages, or it simply times out and does nothing.

I hear there are capacity constraints on Claude. Nothing a few dozen data centers and giga watts of electricity can’t fix.

I have used it (Claude) for financial stuff but found the calcs wrong. Nothing terrible but off a little. I need to go back into it and dig a little deeper into the methodology it is using.

 

[Dave N]

He isn't a troublemaker—he's forum royalty playing a joke on anyone who takes the internet too seriously!

King Troll!

 

[Exhausted]

I’ve struggled with sleep for a few years now and I’ve been using it to evaluate labs, supplements, etc. Just ordered an at home sleep study and plan on using chat to analyze before I step back into an ent’s office again. It does a pretty good job diagnosing symptoms.

Work wise it used to just get me close with python scripting, but it does pretty good now. It just keeps getting better. It’s very useful for importing and exporting CSV files to different formats and comparing the differences.

We’ve used it to diagnose car problems and it’s been spot on.

I have a specific project dedicated to my health as a standalone item. I upload all my doctors notes, all my blood work, and occasionally my body stats from my scale. It provides some interesting insight. And you can tailor it to guide you a particular direction, give critical feedback, or tell you how to stay in line with your goals.

 

[INMT]

Call me a curmudgeon too @Randy11. I don’t even click on that shit on an internet search. Let alone use one on purpose. I even write my own emails for work.

Glad digging in the dirt, building bridges and clearing trail don’t require AI.

 

[Pucky Freak]

Failures of free Gemini:

-confidently states as fact if 51% or more of the information available.

-sycophantic to the point of fueling delusions about user ideas

-Reinforces original response even if it is dead wrong. The more it repeats itself, the more confident it is (stubborn hallucination)

-Fails to adequately warn user at times when their ideas are dangerous or risky. Algorithm is geared towards a warm friendly encouraging personable response over an objective review of reality.

Successes of free Gemini:
-Accurately predicted the # of first choice applicants for a tag I applied for.

-Plumbs the depths of the universe. Like talking to the most knowledgeable person in history (who also happens to be wrong 10% of the time)

-Dramatically speeds up car repairs

-Optimized my closet space in a way I’d never imagine on my own.

-Creates detailed plan outlines to perform complex tasks of any kind

-Cuts reading long emails into 1/4 size
For easy read. Edits my emails for clarity and brevity

- Excellent companion for subtle food prep alterations in real time.

-Optimized major tax reduct/savings over a span of years. I got the IRS to levy a fee ON ITSELF of $368, paid to me last week -100% legal.

My work recently implemented Gemini Pro. It can handle very complex applied math, ranking weighted probabilities, and offering suggestions for deeper analysis. It really shines when user prompts force the algorithm to approach the problem from different angles.

 

[Pucky Freak]

I also use this instruction prompt to strip away Gemini’s human-like tone, and give brutally critical answers. It sometimes begins to revert to the default algorithm, but you can command it to follow the initial instructions and it will revise its latest response in the thread:

ADOPT THE FOLLOWING SYSTEM INSTRUCTIONS FOR THIS SESSION: [Core Persona: Act as a neutral, high-level dialectical interlocutor. Maintain a "cool" professional distance and an academic tone characterized by advanced sentence structure and intellectual rigor.

The Anti-Validation Rule: Strictly avoid all affirming or "cheerleading" language (e.g., "you're right," "exactly"). Do not reinforce user confidence. Explicitly state all uncertainties, data gaps, or logical inconsistencies in the user’s prompt. If a prompt is ambiguous or contradictory, demand clarification before proceeding.

Dialectical Framework (Sequence of Response):

1. Devil’s Advocate: Provide a firm counter-argument to the user’s primary proposition using objective data.
2. Minority Viewpoint: Introduce a secondary, objectively worthwhile alternative theory or conclusion that differs from the mainstream view.
3. Paradoxical Synthesis: Practice paradoxical thinking by identifying how these conflicting truths coexist. Frame the issues through their specific contexts (Intellectual Relativism).
4. Resolution: Guide the user toward a synthesis of these tensions, or, if inconclusive, leave the "loose ends" on the table.

Knowledge Profile:

• Primary Lenses: Prioritize multi-layered analysis through: Macrobiology, Psychology, Sociology, Ecology, Geopolitics, History, and Applied Numeracy.
• Methodological Constraint: Translate implications of physics, chemistry, or advanced applied mathematics into the conceptual language of the "Primary Lenses" above.]

Acknowledge receipt of these instructions with a brief, cold, professional confirmation, then wait for my first proposition.

 

[Jacclynn]

If you want to see where AI actually gets crazy useful for hunting, stop asking it basic questions and start feeding it raw data

Last season, I exported a CSV spreadsheet of my trail cam timestamps, combined it with historical hourly weather data (temp, wind direction, barometric pressure), and had ChatGPT run a correlation analysis. It gave me a breakdown of exactly which wind directions and pressure drops triggered daytime movement on specific ridges. It basically built a custom movement predictor for my specific spot.

As for Claude vs ChatGPT, Claude is a beast if u throw massive PDFs at it. If you dump a 150 page state harvest report into Claude it can instantly pinpoint the exact game units that have the highest success rates with the lowest hunter density. ChatGPT is usually better at the spreadsheet and math side, while Claude handles deep text analysis way better.